Recurrent distributed constraint satisfaction via genetic and evolutionary societies of hill-climbers

In this paper, we compare genetic and evolutionary societies of hill-climbers (GSoHCs and ESoHCs) by G. Dozier and V. Rupela (2002) for solving recurrent DisCSPs. Our results show that: (1) the GSoHCs outperform the ESoHCs; (2) the GSoHCs and ESoHCs that use virtual constraints have better performance than those that do not; (3) GsoHCs and ESoHCs which discover multiple solutions per occasion have better performance than those that discover fewer solutions per occasion; and (4) thrashing that results from a dynamic virtual constraint set may adversely affect the performance of genetic and evolutionary search on recurrent DisACSPs.